Fake News Identification using Hybrid Detection Models

"Fake News refers to fabricated news articles created to promote specific agendas or viewpoints". The rapid dissemination of misinformation is facilitated by "across the broad" usage of various social networking platforms. The identification of fake news has become crucial due to its negative impact on society. As a result, developing reliable and precise techniques for identifying fake news has become a crucial area of study. Various forms of fake news exist, such as rumors, disinformation, misinformation, hoaxes, and clickbait, each posing distinct obstacles for detection. This study presents three hybrid models for fake news identification: CNN–LSTM, BiLSTM–CNN, and Parallel CNN–LSTM. These architectures were tested on IFND (Indian Fake News Detection) dataset, publicly accessible on Kaggle, which comprises 56,714 news articles. These hybrid models, which integrate the capabilities of convolutional and recurrent neural networks, were able to reach accuracy levels of 93.35%, 93.73%, and 94.09% respectively. The Parallel CNN–LSTM model showed marginally better performance, effectively identifying both local and sequential patterns in news content.